Reversing agent-made changes
I record changes and their inverse operations, then rehearse the reversal procedure before using it. I keep that procedure executable by a person without the agent. Destructive changes may not be reversible without a saved prior state.
How do I reverse changes an agent made?
I record changes and their inverse operations, then rehearse the reversal procedure before using it. I keep that procedure executable by a person without the agent.
Implementation
- Group a run's writes into a single change set with one identifier.
- Record an inverse operation for each write at the time the write is issued.
- Prefer additive changes and feature flags over destructive edits.
- Rehearse the rollback in a staging environment before the change ships.
- Keep the rollback path executable by a human without the agent.
What can go wrong
- A prose-only rollback plan can omit the steps needed to restore the prior state.
- A destructive migration without a saved prior state may not be reversible.
Checks to try
- Execute the rollback in staging and diff the resulting state against the pre-change snapshot.
Questions this pattern answers
How do I reverse changes an agent made?
I record changes and their inverse operations, then rehearse the reversal procedure before using it. I keep that procedure executable by a person without the agent.
What steps does this method use?
Group a run's writes into a single change set with one identifier. Record an inverse operation for each write at the time the write is issued. Prefer additive changes and feature flags over destructive edits. Rehearse the rollback in a staging environment before the change ships. Keep the rollback path executable by a human without the agent.
How can I check the result?
Execute the rollback in staging and diff the resulting state against the pre-change snapshot.
What can go wrong with this approach?
A prose-only rollback plan can omit the steps needed to restore the prior state. A destructive migration without a saved prior state may not be reversible.
Related projects
ClawKeeper →
ClawKeeper checks proposed Claude agent actions against explicit rules before a model call. Autonomous spend defaults to zero, and the database enforces an append-only audit log.
TechTide Harness Kit →
TechTide Harness Kit uses JSON Schema to declare skill permissions, risk tiers, network access, and approval requirements. Automated checks validate declarations and run prompt-injection fixtures.
CipherClaw →
CipherClaw analyzes recorded agent activity without model calls. It builds causal graphs, scores possible causes, and flags failure patterns.
Swarm 357 →
357 Claude agent roles across six business layers, with portable memory, enforced cost caps, a bash policy gate, and durable checkpoints.
Related patterns
Acceptance tests for agent work →
Executable checks test the task and detect regressions.
Evaluation tests for releases →
A fixed evaluation suite checks release requirements.
MCP tool permissions →
The server checks which tools a caller may use.
Recording agent runs →
A shared trace ID connects the records of a run.
I explain this method here. Current service scope, prices, and terms are on TechTide AI.
Systems Decision Audit at TechTide AI→© 2026 Alex Cinovoj · TechTide AI · Columbus, OH